TypeSafe Jev
TypeSafe's System One model: returns choices, scores and yes/no probabilities instead of text.
OpenAI Accredited Partner


Not every step in an agentic workflow needs a model that writes text. Decision models return fast, typed answers: yes or no with a probability, a choice from fixed options or a score. The Automation Group uses them alongside generative models.
Decision models answer bounded questions with typed outcomes and probabilities. Code stays in control; the model supplies the judgment.
TypeSafe's System One model: returns choices, scores and yes/no probabilities instead of text.
In public beta since October 2026: predicates, choices and scores over text and images via a dedicated endpoint.
An open-source, multilingual decision model under the Apache 2.0 licence that you can host yourself.
For explanations, summaries and reasoning a generative model remains the right choice. The craft is in the combination.
Send an incoming request to the right agent, tool or department.
Assess tickets, documents or leads against fixed criteria.
Check a generative agent's output before it moves on.
When confidence is low, hand the decision to a person or a heavier model.
Discover more about our approach and solutions.
Together we look at which steps deserve a decision model and which a generative one.